Prescription drug costs are the fastest rising component of national health expenditures over the last half decade and are projected to significantly outpace other service sectors for at least another decade to come (Hefler, et al. 2003). This trend would be viewed in a more favorable light from a policy perspective if it were known to have resulted in significant cost offsets in other healthcare spending categories. Moreover, such cost offsets could be taken as a marker for real improvements in population health. We propose to conduct an empirical investigation of the cost-offset effects of certain types of prescription coverage chosen by Medicare beneficiaries using data from the Medicare Current Beneficiary Survey (MCBS) for 1997 through 2004. We will estimate the cost-offset effects on hospital spending of four categories of insurance coverage for prescription drugs: employer-sponsored insurance;Medicaid prescription coverage;self-purchased Medigap policies;and other forms of public coverage. We believe that important insights about the potential cost-offset effects of some aspects of the Medicare Part D prescription benefit will be gained from our results, and will help in formulating expectations about the future viability of the program. Given the billions of dollars in projected Part D costs over the coming decade, finding cost-offsets would provide some hope that this expensive benefit would be sustainable in a Medicare program that is fast approaching insolvency. There are two methodological challenges to be confronted in the estimation of distinct cost-offset effects for the various drug coverage types. First, a non-trivial proportion of the person-years in our MCBS sample have no hospital admissions. Therefore, a two-part regression specification is called for. Secondly, the coverage-type variables are potentially endogenous;i.e. there exist unobservable variables that are both correlated with individual's coverage choice and exert an influence on his level of hospital expenditure. We find that extant econometric methods designed to simultaneously deal with both of these complicating features are either flawed or not suitable in the present context. Therefore, as part of the proposed project, we will develop, test, compare, and program (in Stata(r)) a new unbiased and robust estimator. The proposed research has the following specific aims: 1) to estimate and statistically test the cost-offset effects of the various types of drug coverage;2) to develop a model, estimator, and corresponding STATA(r) software to take account of the potential endogeneity of multiple dummy variables in a conventional two-part modeling context;3) to evaluate the accuracy (unbiasedness) and precision (statistical efficiency) of the most general version of our new estimator using simulated data.